Neuronal Activity in Human Lateral Temporal Cortex ... - CiteSeerX

The Journal

Neuronal Retrieval Michael

Activity in Human Lateral Temporal from Short-Term Memory

M. Haglund,

George

A. Ojemann,

Theodore

W. Schwartz,

and

of Neuroscience,

March

1994,

14(3):

1507-I

515

Cortex during Serial Ettore

Lettich

Department of Neurological Surgery, University of Washington, Seattle, Washington 98195

Neuronal activity was recorded extracellularly from 20 populations in the lateral cortex of the left anterior temporal lobe of 11 patients undergoing awake craniotomy for epilepsy, during an input-distraction-retrieval measure of recent verbal memory that also included two later successive retrievals of the same information after additional distracting tasks. Changes in activity were determined for each 1 set epoch in three major comparisons: (1) the same visual cues used for naming an input to recent memory, naming without a memory component, and a spatial matching task; (2) memory input (Ml), distraction (S), and initial cued retrieval (Rl) from memory, where object naming was the input to memory and naming of other objects the distracters; (3) initial retrieval (Rl) and the two subsequent serial retrievals of the same information (R2, R3). Control comparisons were also made with serial naming and viewing of blank slides, and repeated naming of the same objects. In comparison 1, 13 of the 20 populations showed consistently increased activity during memory input (“memory units”); two others showed changes during language measures. In comparison 2, a significant proportion of all 20 populations, and the 13 memory units considered alone showed increased activity in initial epochs of MI and Rl, confirming earlier findings of increased lateral temporal neuronal activity at memory entry and initial retrieval. In comparison 3, a significant proportion of the memory units showed increased activity in early epochs of Rl and decreased activity in late epochs of R3. This decrease in populations with increased activity at Rl was also evident when Rl was compared to R2 or R2 to R3. Control comparisons showed no evidence of general habituation or decline in activity. Increased neuronal activity occurs in many left temporal neocortical neurons with input and the initial retrieval of an item from recent verbal memory, and the activity fades rapidly with repeated retrievals. [Key words: human, temporal cortex, temporal lobe, recent verbal memory, short-term memory, microelectrode, singleunit recordings, awake craniotomy] An important role for the temporal lobe in recent memory is generally accepted (Milner, 1971). This role has been further defined in man as one for memory for explicit material that Received Feb. 2, 1993; revised July 13, 1993; accepted Aug. 26, 1993. This work was supported by NIH Grant NS2 1724 and the American Association of Neurological Surgeons Research Foundation. We thank Doug Kalk for computer support/programming and Susan Perkins for secretarial assistance. Correspondence should be addressed to George A. Ojemann, M.D., Professor of Neurological Surgery, Department of Neurological Surgery, RI-20, University of Washington, Seattle, WA 98 195. Copyright 0 1994 Society for Neuroscience 0270-6474/94/141507-09$05.00/O

must be retained during a distraction produced by similar material (Squire, 1987), with this type of memory for verbal material related to the left temporal lobe and that for visuospatial material to the right (Milner, 1971). Usually the medial structures of the temporal lobe are assigned this role, especially the hippocampus. However, there is also evidence for lateral temporal neocortex involvement in recent memory derived from effect of lesions in primates (Fuster et al., 198 1) and humans (Milner, 1967; Ojemann and Dodrill, 1987), stimulation mapping in humans (Fedio and Van Buren, 1974; Ojemann, 1978, 1983; Ojemann and Dodrill, 1985), and recording of changes in activity of neurons in lateral temporal cortex during memory measures in primates (Fuster and Jervey, 1982) and humans (Ojemann et al., 1988, 1990). The one study of neuronal activity in lateral neocortex of human left temporal lobe during a recent verbal memory measure frequently recorded increased activity when information entered memory and when it was retrieved but not throughout the period during which it was stored (Ojemann et al., 1988). That study investigated changes in neuronal activity only during a single initial retrieval of an item stored in recent verbal memory. The present study was directed at two issues related to this left temporal neocortical role in recent verbal memory: first, replication of the changes in neuronal activity in another series of patients, using a measure of recent verbal memory with improved internal behavioral controls so that the same function, naming of different objects, is used as the input and output from memory and the distracting task when the memory must be stored, and second, determination of the effect of serial retrieval of the same item on the increased neuronal activity at initial retrieval. The findings reported provide further insight into the neuronal substrate of human recent verbal memory. A preliminary presentation of some ofthese data has appeared in abstract form (Haglund et al., 1990). Materials and Methods Subjects. The subjects included 11 adults on antiepileptic medications undergoing craniotomies for intractable epilepsy under local anesthesia (0.5% Lidocaine and 0.25% Bupivacaine). All patients were left hemisphere dominant for language as determined by preoperative intracarotid amytal perfusion (Wada and Rasmussen, 1960). The project and consent form were reviewed and approved by the University of Washington Human Subjects Review Committee. All patients underwent separate informed consent for the surgical procedure and the experimental testing. Microelectrode recording. The cortical sites for the microelectrode recordings were all within the boundaries of the planned cortical excision. The sites were in the superior and middle temporal gyri, in the first 55 mm from the tip of the temporal lobe. The local anesthetic and 1 cc of Innovar (fentanyl and droperidol) were given 2-3 hr prior to the language testing. The Trent Wells hydraulic micromanipulator was

1508

Haglund

et al. - Neuronal

Activity

during

Memory

A

‘hat”

$4 Distracters

J/ ‘rake”

R

SPATIAL ‘Control’

‘snake’

‘pea< LRetrieval #2 1

@, d Distracters

RZL

1OCk”

I ‘pear’ I- Aetieval #3

I

LANGUAGE ‘Control’

) Angle

tTm3Iii

OvertlSiktt

Naming

Figure 1. Single trials of the behavioral measures used in this study. A, Memory measure. Each trial consists of the 14 illustrated slides. Below are indicated patient responses and the phase of memory represented by each slide. A subsequent trial of this measure, using different object pictures immediately follows. B, “Control” measures: left, one trial of angle matching; right, one trial of naming. Naming was presented twice, once with the instruction to name overtly, the other with the instruction to name silently. Note that physically identical slides are used for memory input, the first slide of the angle matching pair and naming. Identical overt responses are elicited for memory input and overt naming (and no responses with silent naming and the first slide of the angle matching pair).

mounted on the skull clamp post with a small footplate to stabilize the cortex. The footplate did not blanch the underlying blood vessels. The microelectrodes were primarily stereotrodes that were epoxylitecoated tungsten electrodes electrolytically sharpened to diameters of l5 pm (McNaughton et al., 1983; Cawthon et al., 1989). Typical input impedance of the microelectrodes was 3-7 MQ. Microelectrodes were advanced by the micromanipulator to depths of l-3 mm. Recordings were checked for lack of injury discharge or epileptiform burst activity but were otherwise unselected. Once stable recordings were obtained, the behavioral tasks were begun. Recordings where activity was lost before all behavior tasks were completed were not included in this analysis. Behavioral tusks. The behavioral tasks were presented as slides, each shown for 4 sec. The memory measures consisted of trials each containing 14 slides (Fig. 1A). Each trial began with a blue blank slide, following which the patient viewed an object picture to be named (memory input, MI). Three intervening distraction slides were then viewed. These consisted of two objects to be named and an intervening blank slide. These distractor slides all differed from the object to be retained in memory. In no case did the objects to be named have the word spelled out below the object. Thus, the task involved only object naming. The distractor slides were followed by a slide with the word “recall.” This slide was a cue for the patient to recall aloud (Rl) the object name that had been pictured on the slide named after the blue slide, and retained in memory for the 12 set occupied with distractor slides. The serial retrievals were obtained by continuing in the same trial with three more distractor slides (two with different objects to be named and an intervening blank slide) followed by another slide with “recall” where the initial object-naming slide was to be recalled (R2). This sequence

was then repeated a third time (with different objects as distracters) so that a total of three serial retrievals were obtained for each trial (Rl, R2, R3). In each individual trial, the specific object to be retained in memory was not one of the distractor slides. The patients were briefly trained on the paradigm the evening prior to surgery. Because of time constraints in the operating room, only four to eight trials ofthis measure were obtained for each patient. Prior to or following the memory task, the patients performed language tasks where the same slides used for memory inputs and distractors were named overtly and silently in a sequence of 6-l 2 slides (Fig. IB). The patient also performed a spatial matching task using the same object slides (in all tasks these slides had a red stripe across the object) where a red stripe over the object was matched to a following slide with only a red stripe at the same or different angles. The patient stated “yes” or “no” whether the stripes matched or not. Angle matching is a function that has been related to the nondominant hemisphere based on the effects of lesions (Benton et al., 1975), and likely requires a similar level of attention as does the memory input part of the memory measure. The language tasks were always done just before the spatial matching task. However, all these tasks were done after the memory measure for 14 of the neuronal populations, before the memory measure for two neuronal populations, and between trials of the memory measure for four neuronal populations. The sequence of naming and spatial matching tasks to the memory measure had no apparent effect on the classification of a neuronal population as related to memory, or changes with serial retrieval. Neuronal population analysis. Neuronal activity from the microelectrodes, slide markers, and verbal responses were stored for later analysis on FM tape (frequency response, 100 Hz to 6.5 kHz). Using a Macintosh

The Journal

of Neuroscience,

March

1994,

14(3)

1509

Table 1. Changes in neuronal activity on the three major comparisons of this study

Type

Pt./unit

Comparison 1: Mem./nonmem.

Type 1

9123/2 8931/l 9117/l 8869/2 9117/2 9006/ 1 9032/l 9102/l 893 l/2 8904/2 895911 8904/l 886911 9123/l 9006/3 8942/ 1 8942/2 900612 8947/l 9123/3

mi+2; nL-2,+4 mi+2 mi-1,3 mi+2 mi+2; ON- 1 mi+l,2,3;oN-l;nL-4 mi+l,4 mi+l;oN-1 ON-~; nL- 1,2 mi+ 1,2, 3; ON- 1, 3 ON-l;nL-1 mi+l mi-1,2 oN+l, 2 oN+l 0 0 0 0 0

Type 2 Type 3

Type 4

Type 5 Type 6

Comparison 2 Mem. input Rl

Comparison 3: Rl-R3 Location

mi+l, 2,4 mi+ 1, 2, 3 mi+l, 2 mi+l, 2 mi+1,2 mi+ 1,2,4 mi+l, 2 mi+l mi+l mi+l 0 0 0 0 0 0 0 mi+3,4 mi-2,4 0

fade fade fade fade fade fade 0 0 0 0 fade fade fade 0 0 fade 0 0 0 0

Rl+l Rl+l Rl+l Rl+l 0 0 Rl-3,4 0 0 0 Rl+l, 2 Rl+l Rl+l, 3 0 0 0 0 Rl-4 0 R 1+ 2 (speech)

STG MTG MTG MTG MTG MTG MTG MTG MTG MTG STG MTG MTG STG MTG MTG MTG MTG STG STG

Distal pole Depth Other (mm) (mm) behavior 55

40 45 40 45 30 4.5 53 40 35 35 35 40 55 30 30 30 30 23 55

2 2 3 2 3 3 1 1.5 2 1.5 1.5 1.5 2 2 3 3 3 3 1 2

0 ON-1 0 0 0 N/A 0 N/A RL+2 RL+2 0 0 0 0 N/A RL+2 sR-1 N/A N/A 0

Number of trials 6 4 8 6 8 6 5 8 4 6 6 6 6 6 6 4 4 6 4 6

Neuronal populations are separated by type as defined in text. “Comparison I” is between activity recorded during input to memory (mi), overt naming (ON), or matching a line (nL) all in response to physically identical visual cues; l-4 are 1 set epochs of each task, +, increased activity; -, decreased activity. 0, no significant changes.Only changes with p _c 0.0625 are included. “Comparison 2” is between phasesof the memory measure: significant changes are indicated for input (mi), and initial retrieval (RI). “Comparison 3” is between first and third retrieval of same item of information. “Fade” indicates significantly less activity during R3 than RI. “Location”: STG, superior temporal gyms; MTG, middle temporal gyrus.“Distance from pole” is of temporal lobe, “depth” is below cortical surface.“Other behavior” indicates significant (p < 0.05) changes for this population on comparison of overt to silent naming or word reading and analogous matching tasks in the study of Schwartz et al. (1993). N/A, complete data for these comparisons not obtained.

11x, the analog signals from the FM magnetic tape were digitized at 5000 Hz per channel. Neuronal populations were identified by amplitude discrimination and the negative and positive deflections of each wave. Amplitude-frequency histograms were constructed to separate recordings into populations based on amplitude. Recordings obtained during each slide presentation were divided into four equal epochs of 1.0 sec. The second epoch (1.0-1.99 set) always contained the patient’s response during the overt naming of objects (memory input, overt naming, recalls, etc.), and the same epochs were then used for the silent and line-matching tasks. The first epoch included perception of the visual stimulus and internal processing; the second epoch, later or delayed internal processing and any overt output, and the third and fourth epochs, the pause between tasks. The following five comparisons were made for each neuronal population. (1) Changes in neuronal activity during physically identical slides. Physically identical slides were presented as either an input to memory (MI), an object to be named but not retained in memory, or the first slide of a spatial matching pair. Significant changes in activity during memory input on this comparison (or significant decrease in both naming and spatial tasks) would relate a population to recent verbal memory. (2) Changes in neuronal activity during the memory measure. Averaged across all trials, the same verbal output-enunciating an object name-occurs with memory input, two of the three distracting slides and the initial retrieval (Rl), and physically identical slides have been presented for memory input and two of the three distracting slides. Thus, significant changes on this comparison that do not follow these patterns cannot be related to perception or motor output, providing additional evidence of the specificity of changes to recent memory and identifying where in recent memory-MI, storage, or RI-those changes occur. (3) Changes in neuronal activity during serial retrievals of the same items. As the cue to retrieval for R 1. R2. and R3 and the verbal output are the same, changes in neuronal activity indicate an effect of serial retrieval after an additional distraction period. (4) Changes in neuronal activity during the distractor-naming slide

immediately before RI, R2, and R3. Comparison between these three slides provides controls for the serial retrieval comparisons for several nonspecific effects, such as a general habituation or overall decline in activity. (5) Changes in neuronal activity during the blank slides before RI. R2, and R3. Comparison between these three slides provides another control for nonspecific effects, such as a general habituation or overall decline in activity. In the 20 neuronal populations, there was no evidence of any general habituation during individual trials. Statistical analyses. Neuronal activity for all epochs of each slide in each of the five comparisons described above was ranked for each trial, and changes from the mean rank for each epoch evaluated with the Wilcoxon signed rank test. Because of the limited time in the operating room, in some cases it was only possible to complete four trials of the rather long memory measure in the allotted time. If all four trials changed in one direction, either increase or decrease, the Wilcoxon test for an epoch of the comparisons including that test has a p value of 0.062, slightly above the standard 0.05 (Marascuilo and McSweeney, 1977). Of the 20 neuronal populations analyzed, five included four trials of the memory measure, one included five trials, 11 included six trials, and three had eight trials. For those neuronal populations with only four trials, statements at the usual level of statistical significance for an epoch in a specific population were not possible. However, a very conservative statistical statement on changes in a specific epoch of a particular comparison for all populations considered together can be made by using the binomial expansion for 8, 12, or 20 epochs, at p = 0.062 for 20 populations, with significance set at 0.05, after a Bonferroni correction for the multiple tests in each comparison. For example, an epoch of the first comparison described above (the physically identical slides used as memory input, overt naming or first slide of the spatial matching task) would show an overall significant change, if four or more of the 20 populations showed an increase at p 5 0.062, since the binomial expansion at this p indicates that the chance occurrence of four or more events in the same direction in one of the 12 epochs of this comparison is 0.004, which is below the 0.05

1510

Haglund

et al. - Neuronal

Activity

during

Figure 2. A, Proportion of all 20 neuronal populations with changes in activity at R 5 0.062 for each 1 set eDOch of comparison 2, between input,‘distractor, and Rl phases of the memory measure. Horizontal broken line indicates number of populations (5) that might be expected by chance (based on binomial expansion for 20 tests at p = 0.05 with Bonferroni correction). More populations demonstrated increased activity with epochs 1 and 2 of memory input and epoch 1 of initial retrieval than expected by chance. B, Same as for A, but for comparison 3, between the three serial retrievals. Horizontal broken line is number of populations (4) that might be expected by chance for this many tests. More populations than expected by chance with increased activity in epochs 1 and 2 of the initial retrieval are again evident, and this increased activity disappeared with serial retrievals, so by epoch 3 of the third retrieval there are significantly more populations with decreased activity.

Memory

B

retriaval

#l

retrieval

#2

retrieval

13

g t6 1 -1 4 I

$ t2 2. f 0 1 t 4 I 4 4 -T---w------

level after the Bonferroni correction for the 12 tests. This method allows conventional statistical statements of significance to be made concerning the population of neurons, but not the individual neuronal recordings.

The 13 remaining neuronal populations (65%) showedsignificant increaseswith memory input compared to the language and spatial matching tasks(Table 1).

Results Neuronal populations Twenty stable neuronal populations, free of injury discharge, bursting activity, pulsationartifact, or electrodemovement, were recorded from 11 patients. All recordingswere made in lateral temporal neocortex (superior and middle temporal gyri). All recordingswere made in areasthat had shown little or no evidence of epileptiform activity during 3045 min of electrocorticography. The majority of the recordings were made in the middle temporal gyrus (n = 15) with the remainder in the superior temporal gyrus (n = 5). Recordingdepths varied between 1and 3 mm below the pial surface.Usingthe proceduresalready described, single populations were identified at four recording sites; two populations at five recording sites; and three populations at two sites. At those recording sites where more than one population wasidentified, the neuronal populations always showeddifferencesin their firing frequency, amplitude, and behavioral responses.The averagebaselinefiring rate wastypically low for theselateral temporal cortex neurons(< 1 Hz); however, someneurons showedincreasesin activity up to 30 Hz during specific behavioral measures.

Comparison2: relation betweenmemory input, storage, and RI A significant proportion of all 20 neuronal populations show changesduring initial epochsof memory input (MI) and/or the initial retrieval, and not during the distraction period (Fig. 2A). When the 13 neuronal populations related to memory by comparison 1 are consideredalone, 10 (77%) had increasesduring the first 1 set epoch of memory input. The probability of this many or more populations showing increasesduring the first epoch of the memory input is p -C0.000 1 (binomial expansion for p = 0.062 and n = 13). Sustainedincreases,defined as increasesin neuronal activity beyond the first epoch, were found in 60% (6 of 10) of these neurons (p -C 0.0001). Sustained increasesin the neuronalactivity lastinginto the third and fourth epochs occurred in three populations (30%). In these populations, increasedactivity continues beyond the epoch related to overt responses. Initial retrieval (Rl) evoked significant changesin the neuronal activity in seven (53%) of these 13 neuronal populations. Increaseswere predominantly in the first epoch of Rl when the subjectwaspreparingto retrieve the memory input (p < 0.00 1). The significant changesin both MI and Rl were predominantly in the first two epochs, indicating a relationship to the task. These findings for the memory input and Rl contrast with activity during the distraction where the only significant change (p < 0.05) was a decreasein activity in the last epoch of the secondovert-naming distractor slide. When changesin activity

Comparison 1: relation to memory input, language,or spatial matching Of the 20 neuronal populations, five (20%) showed no change related to any behavior on this comparison.Two (10%) showed changesonly to languagetasks(overt naming or silent naming).

The Journal

of an epoch of individual populations that differ from surrounding epochs at p 5 0.0625 are considered, four populations demonstrated increased activity in both MI and Rl (3 1%) six populations only MI (46%) and three only Rl (23%). This comparison within the memory task is potentially independent of comparison 1, that established a population as related to memory. Indeed, three of the seven neuronal populations not related to memory on that initial comparison showed some changes at the p 5 0.0625 level on the comparison within the memory task. However, two of the changes were in a pattern rarely seen with memory neurons: one with decreased activity at memory input, the other with only very late changes (epoch 3 or 4). The remaining “nonmemory” population showed increased activity in the early epochs of initial retrieval, probably representing another population with only R 1 changes. Comparison 3: changes in neuronal activity during serial retrievals Changes in activity during serial retrievals were then determined for all neuronal populations (Fig. 2B) and those related to memory in comparison 1. The majority of neuronal populations related to memory (69%) decreased activity from the first to the third retrieval. In R 1, the majority of changes (15 of l6)‘are in the positive direction while in R2 the percentage of positive changes decreases (5 of lo), decreasing further in R3 (5 of 16). This fading of neuronal activity is a progressive process as indicated by the comparisons of Rl to R2 and R2 to R3 (Fig. 3). The probability of finding seven increases in the first epoch of Rl compared to six decreases of R2 is very low (both p < 0.0000 1). Similarly, the findings of five increases in the second epoch of R2 (p < 0.000 1) and five decreases in the third epoch of R3 (p < 0.00 1) also point toward a progressive fading of the neuronal activity between later retrievals. The early increases appear to be diminished first, since in the R2-R3 comparison the fading occurs during the later epochs. Fading with serial retrieval was a characteristic of all seven memory populations with increased activity at initial retrieval, including all populations with increased activity on both memory input and the initial retrieval. The R 1, R2, R3 comparison is potentially independent of other measures, so changes could be present in populations showing relations to memory input, or even those not related to memory. Two (of six) populations with only increased activity at memory input showed fading, as did one of the seven “nonmemory” populations. Comparisons 4 and 5: changes in neuronal activity during the distractor-naming and blank slides immediately before RI, R2, and R3 The fading of neuronal activity between R l-R2-R3 could represent a general trend for the neuronal activity to decrease during an individual trial. If this hypothesis is correct, there should be a general decrease in activity in the overt-naming slide immediately preceding the R 1, R2, and R3 slides. In Figure 4, this comparison of the number of populations with changes at p 5 0.062 in the three overt-naming distractor slides shows no significant epochs. As previously mentioned, the comparison of the blank slides between MI-Rl, Rl-R2, and R2-R3 also showed no significant changes in neuronal activity during any of the epochs. These findings show that there is not a general trend for neuronal activity to fade during the internal controls; rather, the fading of neuronal activity is specific to the serial retrievals.

A

B

of Neuroscience,

retrieval

tl

retrieval

+k2

March

retrieval

retrieval

1994,

14(3)

1511

iF2

$3

li% !z +2----l-

3. Changes in activity betweenfirst andsecondretrievalsfrom memory(A) and secondand third retrievals(B) for the 13 neuronal populationsrelatedto memoryby comparison1.Brokenlineisnumber of populationsexpectedto showchanges on a chancebasisat p 5 0.05. Significantlymorepopulationsthanexpectedby chancehaveincreased activity in epoch1 of initial retrieval comparedto decreased activity in epoch1 of the secondretrieval, whilewhensecondis comparedto third retrieval, a significantnumberof populationswith increased activity are identifiedin secondand third epochof the secondretrieval comparedto decreased activity in thethird epochof the third retrieval. Thesefindingsindicatea progressive fadingof increased activity with serialretrievalof the sameinformation. Figure

Changesin neuronal activity related to the familiarity of objects Another possibleexplanation for the fading of neuronal activity during the serial retrievals involves habituation becauseof the familiarity of objects being presented.Someof the object slides that were usedasinput to memory were also usedasdistractor slideson other trials. Memory input slidesthat demonstrated fading of neuronal activity during serial retrievals, when used in separatetrials as distracters, were separatedby 6-17 slides and repeatedone to four times. A significant increasein the first epoch of the initial presentation of an object compared to the first epoch of the last presentationwould suggestthat decreased neuronal activity during serial retrievals is due to familiarity of the object beingnamed.No significant changeswere seen.There is also no difference when thoseneuronal populations with fad-

1512

Haglund

et al. * Neuronal

Activity

during

Memory

A overt name I1

overt name #2

overt name #3

Z+O ii l4 l $1 L c b .I 3 i* z x4 0

B

m------m--m--

overt name (11

4. Controlfor decliningactivity asan explanationfor effectsillustratedin Figures 2B and 3. Activity in Figure

the naming slide preceding each retrieval iscomnared for all threeretrievals (A), for fir& compared to second retrieval (B), andfor second comparedto third retrieval (C ). No significant

changes arepresent. ing on serial retrieval (Fig. SA) are compared to those that did not fade (Fig. 5B). Indeed, the first presentation actually has more significant epochs in the negative direction than in the positive direction. These negative findings, in addition to the negative finding for a generaltrend toward decreasingneuronal activity in individual trials, emphasize that the decreasesin activity during serial retrievals are related to memory. Location of neuronal populations associatedwith memory For the purposesof anatomical localization, the 13 memoryrelated neuronal populations were divided into (1) four populations with increases(at the p I 0.062 level) during MI and Rl with fading during the serial retrievals (type I), (2) two populations with increasesduring MI and fading during serial retrievals (type II), (3) four populations with increasesduring MI without R 1 changesand no fading (type III), and (4) three populations with no changeduring MI but increaseswith Rl and fading during serial retrievals (type IV). The nonmemory populations were divided into language-related(n = 2; type V) and five populations with no changeduring any behavioral measure (type VI) (Table 1). The anatomical location of the neuronal recording siteswere recorded for the gyrus (middle or superior temporal); the distance from the temporal tip (range, 23-55 mm); and the depth below the pial surface(l-3 mm). No major pattern in the anatomical location and type of neuronal population wasobvious, except that the memory-related neuronal populations in the middle temporal gyrus were posterior compared to the more anterior nonmemory neuronal populations (Fig. 6A). The superior temporal gyrus contained no pattern in the anatomical location or in the type of neuronal population.

overt name rt2

C.

overt name #2 * 71

overt name M

I 1* i4 .o 0 .: h 3 -mw------*-t s 4 j4 l

In the comparison of the microelectrode depth and its relationship to the type of memory-related neuronal populations, one significant theme was observed (Fig. 6B). There were no differences between the depth when comparing neurons with changesduring memory input, initial retrieval, or fading during serial retrieval. However, when comparing those memory neurons with increasesin MI and fading of Rl-R3 (neuronal types I, II) to those memory neuronswithout this combination (neuronal types III, IV), a significant differencewasobserved. Those neurons with increasesduring MI and fading during serial retrievals (n = 7, types I, II) were deeper2.5 + 0.2 mm compared to those memory-related neurons(n = 6, types III, IV) without both thesechanges,1.6 + 0.1 mm (p = 0.028). Memory-related neuronsand language Comparison 1 indicated a separation between neurons related to memory and those related to naming. In addition, aspart of another study (Schwartz et al., 1992) many of thesesameneurons were analyzed for changesduring additional behavioral measurescomparing overt to silent word reading, and a wordreading spatial matching task, as well as comparing overt to silent naming and the analogousspatial matching task. Of the memory-related neuronal populations (n = 13), 11 populations were analyzed for these additional behavioral measures.Only three of the 11 neuronal populations (27%) related to memory showany changeduring theseother behavioral measures.In the three neuronal populations that did showchangeswith the other measures,two were increasesin activity during the spatial task in responseto a word with a line acrossit (rather than an object and line) and one was a decreasein activity during the first epoch of the overt naming task. These comparisonsprovide

The Journal of Neuroscience,

first presentation

last oresentation

sylvian

March

superior middle

B

first oresentation d -11 $+4 0 = +2 i! s CJ 0 a 5 -2

1513

,

flssura

0

1994, f4(3)

10 20 30 40 mm from left temporal

50 60 pole

temporal temporal

A= type O=type6

gyrus gyrus

5

last oresentation . . . . . . . .. . .. . . . . . .

5

Figure 5. Control for general habituation as an explanation for effects illustrated in Figures 2B and 3. Activity is compared during an initial and subsequent naming of the same object picture for populations showing fading of activity with serial retrieval (A) and for populations without fading (B). There is no evidence of reduced activity with repeated naming of the same object picture.

further evidence that the memory-related neuronal populations are primarily related to memory and not other languageor spatial behaviors. Of the seven neuronal populations not related to memory, only four were analyzed for the other behavioral measures. However, three of the four (75%) had changesrelated to the other behavioral measures.One unit related to languagein the comparison of this study had increasesin activity in the first epochsof the overt and silent reading tasks.Two units with no behavioral changeson comparison 1 of this study showedreciprocal changesto silent readingand the spatialtask in response to a readingstimulus. Although thesecomparisonshave a small samplesize,a minority of memory-related neuronalpopulations and the majority of nonmemory neuronal populations show changeswith other behavioral measuressuch as naming and reading. Seven patients had microelectrode recordingsseparableinto two or more neuronal populations (Table 1). Three patients, including both patients with three populations, show very different behavioral correlatesin the nearby populations.The other four patients show changeswith the memory measurefor both populations, but no two nearby populations had significant changesin the sameportion of the memory measure. Discussion Changesin neuronal activity during memory input and initial retrieval This study confirms the finding in our earlier study that a large proportion ofneurons sampledfrom left dominant anrerior temporal neocortex increaseactivity with recent verbal memory

- ---

31mm,.,~~ c

white

matter

Figure 6. Location of types of changes in neuronal activity with memory measure. A, Lateral projection. B, Cortical depth. Types of changes in activity are defined in Results.

measures.In the present study we have usedtwo comparisons to relate a population to memory: (1) changeson the memory measurewhen the visual cueswere physically identical for three different tasks-memory input, a languagetask without any memory component, and a spatial matching task-as well as having the same verbal output for the memory and language tasks;and (2) changeswithin the recent verbal memory measure basedon the input-distracter-output paradigm of Petersonand Peterson (1959). In the first comparison, 13 (65%) of the 20 sampled populations were related to memory input by both measures.Our previous study showed a similar proportion of populations related to memory, 12of 17 (70%) (Ojemann et al., 1988). The probability of seeinga relation to memory on both measuresfor an individual population is quite small and suggeststhat a large proportion of neurons in the lateral temporal cortex are dedicatedto processingmemory input and the initial retrieval from memory. The secondcomparison adds another dimension to the understandingof the role of the lateral temporal lobe in memory. Increasesin neuronal activity in the secondcomparison were related to memory input in 77% of the memory-related populations asdefined by the first comparison, and 58% in the previous study (Ojemann et al., 1988). This activity occurs essentially every time new information is encountered, for only that pattern would achieve the significancelevels required to relate a neuronal population to the input memory phase in either study. This increasedactivity when a new item enters recent verbal memory is often relatively long lasting,extending beyond the time required to identify the item verbally in 30% of the neuronal populations with memory input changesin the present study and 50% in the earlier study. Increasedactivity at initial retrieval wasevident in 53% of the memory populations of the presentstudy and 58% of the earlier study. These findings are supported by microelectrode recordings from monkey inferior temporal lobe during a visual delayed

1514

Haglund

et al. - Neuronal

Activity

during

Memory

match-to-sample task. Fuster and Jervey (1982) found a large proportion of neurons with changes in neuronal activity during the sample and match after a 15-20 set delay (sample: changes, 96%; no change, 4%). In contrast, a much smaller percentage of neurons showed changes in the intervening delay between the sample and match (changes, 44%, no change, 56%). This type of change was less evident when the delay was filled with distractors (Fuster, 1990). These findings demonstrate that a large proportion of neurons in the primate inferior temporal region also show changes with visual memory measures. Changes in neuronal activity during serial retrievals The present study indicates that the increase in neuronal activity at the initial retrieval is not sustained through subsequent retrievals of the same item. A reduction is already apparent by the second time an item is retrieved and even more so by the third. The reduction in activity is quite specific to items in recent explicit verbal memory. No overall reductions in neuronal activity occurred nor did reductions occur with serial naming of the same items when they were not to be retained in memory. Thus, there is no evidence from our data for an overall habituation or for effects of novelty, per se. This transient increased activity ofmany temporal neocortical neurons at memory entry is apparently essential for retention of items. The verbal memory deficit after left temporal lobe resections has been related in part to the extent of the neocortical removal (Milner, 1967; Ojemann and Dodrill, 1985). With surface electrical stimulation during memory entry at many sites in left anterior temporal lobe, although the object can be correctly named, it is not retained in recent verbal memory (Ojemann, 1978, 1983). Surface stimulation presumably alters the increased activity at memory entry. The reduction in this activity with serial retrieval occurs in most memory-related temporal neocortical neurons. This suggests that if neuronal activation is part of retaining or retrieving an item that has successfully entered human memory, as is suggested by invertebrate studies of conditioning (Hawkins et al., 1983), those activated neurons are either located elsewhere or are a very small subset of temporal cortical neurons. The widespread initial increase in activity at memory entry (and to a lesser extent at initial retrieval) is perhaps best viewed as a promoter for retention of an item in memory. Damasio has proposed that recall involves simultaneous activation of multiple small cortical “convergence zones” (Damasio, 1990). Perhaps the widespread temporal activity at memory entry sets up the convergence zones for this memory, or the process that “binds” them together. Activation of smaller neuronal populations then would be required for subsequent retrieval, once the binding phenomenon has taken place after memory input and initial retrieval. Although the present study was not specifically a test of learning, correct serial retrieval often leads to learning as evidenced by more accurate later retrieval. Thus, learning of an item of information seems to involve this process of reducing the neuronal activity involved in retrieval ofthat item. Similar evidence has come from both animal and human studies. The activity of primate inferior temporal cortex neurons increased with the sample and/or match in a visual delayed match-to-sample task; however, these changes were not sustained, as further repetitions led to a decrement of the response (Bayliss et al., 1987; Bayliss and Rolls, 1987). Just as the response of temporal lobe neurons to visual memory input and retrieval appear to be widely distributed, the fading of neuronal activity with serial recall of an

item entered into memory appears to be distributed across multiple areas of the primate temporal visual cortex. In human studies, patients with lesions in the temporal lobe show marked impairment of recall in paired-association tasks (Milner, 1967, 197 1). Positron emission tomographic (PET) studies also show a shrinkage in areas of initially increased blood flow during learning of a motor (Seitz et al., 1990) or a semantic task (Raichle, 1990). Changes in neuronal activity during other behavioral measures The present study, like previous studies of human temporal lobe neuronal activity during measures of higher functions (Haglund et al., 1993; G. Ojemann et al., 1988, 1990; J. Ojemann et al., 1992) found considerable specificity to the behavior eliciting changes in activity of a particular neuronal population. Only 2 of the 13 “memory units” showed changes in activity with any other measured behavior that was in the same direction as the memory change, in both cases to the spatial matching task involving reading stimuli. These two memory units showed a reciprocal change in activity between memory and overt naming, and one other memory unit showed reciprocal activity between memory input and overt naming. Thus, most neurons are apparently activated by only one of the class of memory, language, or spatial behaviors. This degree ofspecificity is somewhat surprising, given the large proportion of the sample responding to memory. However, frequent separation between recent verbal memory and language mechanism was also a finding in stimulation mapping studies oftemporal cortex (Ojemann and Dodrill, 1985; Ojemann, 1978, 1983). In our other studies of changes in neuronal activity during language and visuospatial measures (Schwartz et al., 1992; Ojemann et al., 1992) nearby neuronal populations recorded by the same microelectrode usually had different behavioral correlates. In the present study, two-thirds of nearby neuronal recordings responding to memory demonstrated changes to memory in all nearby populations-though in somewhat different ways. No nearby populations altered activity during exactly the same memory phases. Based on the likely consequence of resection, left anterior temporal neocortex is probably more essential for explicit verbal memory than for language or spatial functions. Because of the invasive nature of microelectrode recordings, little is known about the activity in essential language areas. However, the present study suggests that the presence of a high proportion of neurons, including nearby neurons, changing activity with a behavior, may be a feature of cortical areas essential for that behavior. References Bayliss GC, Rolls ET (1987) Responses of neurons in the inferior temporal cortex in short term and serial recognition memory tasks. Exp Brain Res 65:6 14-622. Bayliss GC, Rolls ET, Leonard CM (1987) Functional subdivisions of t:mporal lobe neocortex. J Neurosci 7:330-342. Benton A, Hannay H, Vamey N (1975) Visual perception of line direction in patients with unilateral brain disease. Neurology 25:907910. Cawthon D, Ojemann G, Lettich E, Kalk D, Percival D (1989) Human extracellular stereotrode analysis of neuronal ensembledynamics during languagetasks. Sot Neurosci Abstr 15:302. Damasio AR (1990) Synchronous activation in multiple cortical regions: a mechanism for recall. Semin Neurosci 2:287-296. Fedio P, Van Buren JM (1974) Memory deficits during electrical stimulation of speech cortex in conscious man. Brain Lang 1:29-42. Fuster JM (1990) Neuronal discrimination and short-term memory

The Journal

in association cortex. In: Neurobiology of higher cognitive function (Scheibel A. Wechsler A. eds). v v 85-101. New York: Guilford. Fuster JM, Jervey JP (1982) ‘Neuronal firing in the inferotemporal cortex of the monkey in a visual memory task. J Neurosci 2:361375. Fuster JM, Bauer RH, Jervey JP (198 1) Effects of cooling inferotemporal cortex on performance of visual memory tasks. Exp Neural 7 1: 398-409. Haglund MM, Ojemann G, Cawthon D, Lettich E (1990) Neuronal activity in human temporal lobe with multiple memory retrievals. Sot Neurosci Abstr 16:760. Haglund MM, Ojemann GA, Lettich E, et al. (1993) Dissociation of cortical and single unit activity in spoken and signed languages. Brain Lang 44: 19-27. Hawkins R, Abrams T, Carew T, Kandel E (1983) A cellular mechanism of classical conditioning in Aplysia: activity dependent amplification of presynaptic facilitation. Science 2 19:400-405. Marascuilo L, McSweeney M (1977) Nonparametric and distributionfree methods for the social sciences. Monterey, CA: Brooks/Cole. McNaughton B, O’Keefe J, Barnes C (1983) The stereotrode: a new technique for simultaneous isolation of several single units in the central nervous system from multiple unit records. J Neurosci Methods 8:391-397. Milner B (1967) Basic mechanisms suggested by studies of temporal lobes. In: Basic mechanisms underlying speech and language (Millikan C, Darley F, eds), pp 122-133. New York: Grune and Stratton. Milner B (1971) Disorders of learning and memory after temporal lobe lesions in man. Clin Neurosurg 19:42 l-446. Ojemann GA (1978) Organization of short-term verbal memory in language areas of human cortex: evidence from electrical stimulation. Brain Lang 5:33 l-340. Ojemann GA (1983) Brain organization for language from the perspective of electrical stimulation mapping. Behav Brain Sci 6:189230.

of Neuroscience,

March

1994,

74(3)

1515

Ojemann GA, Dodrill CB (1985) Verbal memory deficits after left temporal lobectomy for epilepsy: mechanism and intraoperative prediction. J Neurosurg 62: 10 l-l 07. Ojemann GA, Dodrill CB (1987) Intraoperative techniques for reducing language and memory deficits with left temporal lobectomy. In: Advances in epileptology, Vol 16 (Wolf P, Dam M, Janz D, Dreifuss F, eds), pp 327-330. New York: Raven. Ojemann GA, Creutzfeldt OD, Lettich E, Haglund MM (1988) Neuronal activity in human lateral temporal cortex related to short-term verbal memory, naming and reading. Brain 111:1383-1403. Ojemann GA, Cawthon DF, Lettich E (1990) Localization of physiological correlates of language and verbal memory in human lateral temporoparietal cortex. In: Neurobiology of higher cognitive function (Scheibel AB, Weschler AF, eds), pp 185-202. New York: Guilford. Ojemann J, Ojemann G, Lettich E (1992) Neuronal activity related to faces and matching in human right nondominant temporal cortex. Brain 115:1-13. Peterson L, Peterson M (1959) Short-term retention of individual verbal items. J Exp Psycho1 58: 193-198. Raichle M (1990) Exploring the mind with dynamic imaging. Semin Neurosci 2:307-3 15. Schwartz T, Ojemann G, Haglund M, Lettich E (1992) Cerebral lateralization of neuronal activity during naming, reading and line-matching. Epilepsia 33[Suppl 31:120. Seitz R, Roland P, Bohm C, Grertz T, Stone-Elander S (1990) Motor learning in man: a positron emission tomographic study. Neuroreport 1: 17-20. Squire L (1987) Memory: neural organization and behavior. In: Handbook of physiology, set 1, The nervous system, Vol V, Higher functions of the brain, Pt 1 (Mountcastle V, Plum F, Geiger S, eds), pp 295-372. Bethesda, MD: American Physiological Society. Wada J, Rasmussen T (1960) Intracarotid injections of sodium amytal for the lateralization of cerebral speech dominance. J Neurosurg 17: 266-282.